Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Sony IMX183 mono test thread - ASI, QHY, etc.

  • Please log in to reply
814 replies to this topic

#1 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 22 November 2017 - 03:44 PM

Hello all. I know some of you have been waiting for my report on the ASI183MM Pro. I've been working with it for a few months now, and have data for several images. So far it's been entirely narrow band, however I have started acquiring LRGB data on M33. I am still evaluating the camera, both from a normal usage standpoint, as well as from a low level characteristic standpoint. So more will come. 

 

I am purposely starting this thread as a generic test thread for Sony IMX183 based cameras. Doesn't matter all that much to me which brand, all of them need testing, and with CMOS sensors so much of the functionality is integrated into the sensor, most of the differences between the cameras are pretty minor (especially these days, as ZWO has caught up to QHY and Atik in terms of standardized camera build and standard features like DDR memory buffer and the like.) So feel free to talk about any brand in this thread, I encourage it. The key is the underlying sensor, the IMX183 mono (there is also a color version, discussion about that sensor is also welcome, and I know there is a QHY183C on the market already, and I think it has been given too little attention...I've tested some data from that camera as well, and I thought it was quite good.) 

 

The Sensor

 

Starting with the sensor itself. This is a new largish monochrome CMOS sensor. It's the second such sensor to find it's way into astro cameras since the ASI1600, which was the first mono CMOS sensor larger than a fractional inch size to be placed into an astro cam. The Sony IMX183CLK is slightly smaller than the Panasonic MN34230ALJ, the latter being a micro 4/3 format sensor (1.33x as long as tall for the aspect ratio), the former being a standard 1" format  (1.5x as long as tall for the 3/2 aspect ratio). The Sony sensor has 2.4 micron pixels, compared to the Panasonic's 3.8 micron pixels, making the Sony pixel area about 2.5x smaller. The Sony sensor has 20 megapixels, compared to the Panasonic's 16 megapixels. Image dimensions are 5496x3672. FITS file size is just about 40 megs.

 

Dark current is very low, possibly lower on the Sony sensor than the Panasonic sensor. Like most CMOS sensors, it has amp glow, and this makes getting an accurate measurement of the dark current of the sensor more difficult, as true dark current generally grows at a different rate than the amp glows grow. Additionally dark current growth is usually linear, while amp glow growth tends to be non-linear as well as non-uniform across the sensor. Testing is still being performed to get a more accurate read of the real world dark current for this sensor, as well as to characterize the dark current separately from the amp glow, at least to some degree.

 

The rated dark current rate for the IMX183CLK is ~0.002e-/s @ -20C. In comparison to the MN34230ALJ which is rated at ~0.006e-/s @ -20C. Read noise for the IMX183CLK ranges from ~2.9e- at minimum gain (0), to as little as ~1.4e- at maximum gain (270, asi). This is a similar range to the MN34230ALJ, which has read noise ranging from ~3.5e- at minimum gain (0), to as little as 1.1e- at maximum gain (300, asi). 

 

Initial Conclusion

 

With such small pixels, but a reasonably largish sensor, it puts the IMX183 based cameras in an interesting position for high resolution imaging. The pixels are the smallest you can get right now in reasonably sized sensors, making it the highest resolution sensor on the market alongside the IMX178. This has interesting implications for both high resolution imaging of smaller objects such as galaxies, but at more moderate focal lengths of around 800-1000mm, as well as for better-sampled imaging of larger objects, nebula, molecular clouds, etc. at shorter focal lengths of 135-400mm. One of the toughest things about wide field imaging is poor image scale, which leads to very bloated stars that can often be difficult to manage. The IMX183 may shine most as a high resolution wide field imager, and in general as a decent high resolution imager for efficient, smaller and more cost effective telescopes (i.e. instead of needing a C14Edge with a KAF-16803 to get decent high resolution images, you could use a much more cost effective 8" f/4 newt, and get similar results to the SCT/KAF setup, but in less time). 

 

Read Noise

 

On the read noise front, it is important to note that these are the per-pixel values. Because the IMX183 has smaller pixels, the read noise in apples-to-apples terms is actually higher than that of the MN34230ALJ. When normalizing the pixel size to 3.8 microns, the IMX183 has an effective max read noise of about 4.5e-/3.8um, and effective minimum read noise of about 2.2e-/3.8um. 

 

This was a slightly surprising discovery to me. My original expectations of the ASI183MM Pro were that it would have a 14-bit ADC, much like the ASI178 which has similar pixel size, as well as similar read noise to the ASI178. The ASI178 has read noise ranging from about 2.25e- down to about 1.3e-. The ASI183MM Pro ended up having a 12-bit ADC, which adds some additional quantization error, which is another noise term rolled into the total read noise, and generally accounts for the differences. However, with the smaller pixels, it does make the camera a slightly higher noise camera, in line with the newer Sony ICX CCD sensors. 

 

Due to the higher read noise, the IMX183 benefits from longer exposures than are necessary with the MN34230ALJ. This is not an issue with LRGB, and even with exposures doubled or tripled compared to the MN34230ALJ, exposure times are still entirely reasonable. For some people, the longer LRGB exposures may even be much preferred, for those who don't like short exposures or don't want to stack hundreds of LRGB subs. This should make the IMX183 cameras a very attractive option for galaxy imagers looking to maximize resolution on a more reasonable budget. I think an 8-10" f/4 newt, or refractors in the 130-180mm f/6-f/7 range, paired with the IMX183 Mono could deliver amazing results for galaxies. 

 

Narrow band is a slightly different story. Where the MN34230ALJ can handle 3-5 minute subs at a higher gain and effectively swamp the read noise, the IMX183 has more limited DR at high gain, and swamping the read noise becomes more of a challenge. In my testing with narrow band, I have found that gain 53 (asi) combined with 7-10 minute subs at f/4 delivers better results for narrow band. The SNR per sub is better, DR is quite good. This is again in line with Sony ICX CCD sensors which have similar amounts of read noise in a normalized 3.8 micron effective pixel. One might think of the IMX183 as the closest CMOS counterpart to a Sony ICX CCD, in my experience the results are very similar (main comparisons have been ICX814 data I've worked with in recent months.) 

 

YUO9vr6.jpg
Calibrated IMX183 300s dark frame, starburst glow area, to show noise

 

Amp Glow

 

Like most consumer grade CMOS sensors, the IMX183 has amp glow. This shouldn't be surprising to anyone, however it may be interesting to know the differences and similarities with other sensors. The IMX183 is a Sony sensor, unlike the MN34230ALJ which is a Panasonic sensor. Sony sensors have a fairly consistent amp glow characteristic, an the IMX183 shares it.

 

The glow characteristic is two radial glows in the lower left and right corners, small, covering only the corners, plus a starburst glow (a multi-rayed glow that sends rays off across the area of the sensor) off to the right edge. There is a very slight and large scale glow to the upper left that is often imperceptible outside of deep stacks of darks. This is the same as those familiar with some of the other Sony IMX sensors, such as the IMX178, IMX294, etc. 

 

In contrast, the Panasonic glows are a very faint large scale glow to the upper left, and the double-bubble along the right edge. The double bubble is actually more like a half doughnut shape, once you've stacked a decent number of darks into a master. 

 

There is another glow characteristic difference between the Sony and Panasonic sensors. The Panasonic sensors have a larger read-time dependent factor in how bright the glows get than the Sony sensors, and a smaller exposure-time dependent factor. Sony sensors may have a slight read-time dependency, however they appear to be primarily dependent on the length of the exposure. As exposure time grows, so do the glows. With bright signals, the glows can be effectively buried, so with either camera they are usually a non-issue for LRGB or OSC imaging.

 

With fainter signals and narrow band, the glows of the IMX183 could potentially be a bigger issue. Narrow band imaging is very effective with the IMX183 (most of the images I'll be sharing so far have been narrow band), however with longer exposures of 10 minutes or longer, the amp glows can leave a bit of extra noise in the corners and near the starburst. With deep stacks this isn't an issue, however for mosaicing, it could potentially leave some visible seams in those areas.

 

Calibration 

 

Like any camera that has amp glow, best practice is to calibrate with well matched darks (same gain, offset, temp and exposure time), without any dark optimization/dark scaling. Glows calibrate out perfectly with proper calibration. The hot pixels on the Sony IMX183 appear to be much more stable than the hot pixels on the Panasonic MN34230ALJ. This makes dark frame subtraction far more effective in removing hot pixels with the IMX183 than the MN34230ALJ.

 

The Panasonic sensor appears to have more RTS (random telegraph signal) which leads to semi-fixed hot pixels. RTS noise can result in certain pixels oscillating between two or three (or even more) relatively fixed states. These are not hot pixels in the classic sense, but they look the same in a sub exposure. The problem with RTS pixels is they are inconsistent. The same pixels do not always exhibit the same way in every frame, so even with dark subtraction, many "hot pixels" may remain in the data. This makes dithering doubly important with the Panasonic sensor. Additionally, it can make Cosmetic Correction a very useful step, but that adds more work to the pre-processing workflow.

 

The IMX183 does not appear to need any additional cosmetic correction step after simple dark subtraction. After dark subtraction, I actually find the noise characteristic of the IMX183 is extremely pleasing, very clean, cleaner even than the Panasonic MN34230ALJ. It has a very CCD-like quality to it once calibrated, which should please many people. There is some slight banding that can appear at lower gain settings, so dithering is still highly recommended. In general, I recommend dithering with all cameras, regardless of the sensor type or usage pattern. FPN can exhibit in many ways, even spatially random noise, and dithering will always benefit the final integrations by ensuring that any patterns are randomized through the stack. Utilize dithering with the IMX183 as you would any other camera.

 

Release Dates and Scope Pairings

 

I don't have any concrete information on release dates for these cameras. What I have heard is they are coming soon, though.  End of November to early December, and they should be hitting the street. From both ASI and QHY, as I understand, although who will hit the streets first I cannot say. I do not know if Atik is working on a camera with this sensor (I did ask, but they like to keep their cameras under wraps until they can get them out for beta testing), nor do I know if anyone else is working on one. For those interested in one of these cameras, keep an eye out.

 

The IMX183 is a very high resolution sensor. It's got ultra tiny pixels. This means you can get more out of the resolution of any given scope than with most other cameras...but it also assumes you have the seeing to support the kind of resolution this sensor can deliver as well. For those interested in this camera, if you truly have an average of 3" seeing, then you may not benefit from the resolution potential, and you may be better off with a camera using the Panasonic sensor. For those who have around 2" seeing or better, then you should indeed be able to benefit from the resolution this camera has to offer. 

 

3BNPUIx.jpg
Very small ~1.7" FWHM stars on a night of good seeing in Colorado, ASI183MM Pro @ 600mm fl, f/4

 

For telescope pairings. Because of the tiny pixels and 1" sensor size, I would avoid pairing it with larger scopes. There is no need to use an RC or SCT with this sensor. Doing so would result in rather ridiculous image scales of 0.3"/px (8" RC) to as little as 0.12"/px (14" SCT)!! If you live in Paranal, and are slapping one of these cameras on something like an f/2 2-meter aperture scope, then you might be set! ;P However, I recommend shorter/smaller scopes.

 

I think for high resolution LRGB galaxy imaging, the ideal pairings would be with 8" or 10" newtonins, f/4 to f/5. That will deliver an image scale of about 0.5-0.6"/px, which would sample 2" seeing limited stars by 3.3-4x. Right in the sweet spot for optimal high resolution imaging, IMO. The f/4 or f/5 f-ratio should deliver photons quickly, giving you read noise swamping subs in around 60-90 seconds or so for L, and 2-3x that for RGB (gain 53, asi). I also think you could easily pair this with refractors in the 130-150mm aperture range, at f/6-f/7. That will deliver an image scale around 0.51-0.65"/px. A TEC140, for example, is 980mm long at f/7, should make for a pretty awesome scope for high res galaxies if you've got the seeing for it. 

 

For wide field narrow band, I think the ideal focal length range is going to be from around 200mm, up through about 400mm. This will give image scales in the range of ~2.4"/px to around 1.2"/px. For wide field nebular imaging, that is again pretty much the sweet spot. At 2.4"/px, your sampling ratio for 2" seeing would be closer to 1x, which is a lot better than say 5.4 micron pixels which would have a tiny image scale of 5.5"/px @ 200mm! You should be able to avoid terribly blocky stars with around a 1x sampling. At 400mm you would be sampling 2" seeing at around 1.7x, which should ensure you don't have blocky stars. This should also render stars at a more reasonable size, so that with wide fields the stars don't totally dominate the image (often a problem with image scales of 3"/px or smaller.) At such image scales, narrow band imaging becomes a totally viable option, thoroughly swamping the read noise should only require exposures of a few minutes. I'm quite curious to see what kinds of large scale mosaics could be created at 400mm, without being so totally dominated by stars, and with higher resolution and more details...should be quite interesting!!

 

In final conclusion, don't expect the IMX183 to be an "ASI1600 killer" in any way. The IMX183 is actually a slightly higher noise camera. It's a camera with a different purpose. With such tiny pixels, it has exciting implications for high res imaging with much smaller scopes than have classically been required, without the need to spend the big bucks on very large cameras that have very large pixels. With LRGB imaging, read noise won't matter much, however the smaller pixel size could deliver much more detailed images than any other camera at any given focal length (depends on seeing). It could also present some interesting implications for true lucky imaging, separating double stars, lucky imaging of small planetary nebula with a more budget-friendly telescope, etc. 

 

Well, that's it for now. I have some example images to share, some low level dark and bias frames and comparisons to share, and some noise and gain evaluations in the works. I'll post more about those over the coming week. 


Edited by Jon Rista, 22 November 2017 - 04:14 PM.

  • timmbottoni, dswtan, h2ologg and 26 others like this

#2 corsiar

corsiar

    Ranger 4

  • *****
  • Posts: 391
  • Joined: 29 Aug 2015
  • Loc: Tacoma, WA

Posted 22 November 2017 - 07:07 PM

I had a OSC of this camera for beta testing last year and really liked it but always wanted a mono version and it is here. I want to stick one of these on my GT102 which would be around a 0.7"/pixel. So does my tracking have to be that good or better to get round stars? My average RMS through an OAG is around 0.5-0.75 most nights some night high as 1.0-1.2. Currently using a QHY163M.  



#3 entilza

entilza

    Gemini

  • *****
  • Posts: 3326
  • Joined: 06 Oct 2014
  • Loc: Oakville, ON, Canada

Posted 22 November 2017 - 11:41 PM

A very nice review Jon!  Thanks!   This lens would be quite nice with the Samyang 135mm.  I'll have to wait though as I've just got this camera but I am looking forward to someone trying that combination.  Would give a scale at around 3"  which is quite nice for that field.


  • premk19 likes this

#4 premk19

premk19

    Messenger

  • *****
  • Posts: 492
  • Joined: 19 Apr 2015

Posted 23 November 2017 - 12:41 AM

Thanks, Jon! Looking forward to seeing more reviews and images. Considering this for my 135mm & 180mm lenses and 60mm f/4.3 doublet. May also give it a go with my 115mm f/7 refractor for some galaxy imaging.



#5 alien_atx

alien_atx

    Lift Off

  • -----
  • Posts: 8
  • Joined: 04 Aug 2016

Posted 23 November 2017 - 01:20 PM

Thanks Jon for this analysis!

 

Could you comment on pixel size vs telescope spot size?  I am thinking about pairing a IMX183 cam with a RASA 11 ( 620mm fl and ~2.2 micron spot size ).  The camera pixel size ( 2.4 microns ) almost perfectly matches the telescope spot size ( 2.2 microns ), which could yield some amazing resolution.  I am hoping to use the IMX183 / RASA combo for high resolution planetary nebula imaging.


  • Pug likes this

#6 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 23 November 2017 - 02:01 PM

You guys are welcome. :) I'll be sharing some example images today, then some time this weekend and next week I hope to be sharing at least some examples of dark and bias frames and the like, along with some analyses. I'm working on a PTC analysis since I'm having trouble determining the full noise characteristic jus with PI standard measurements.

 

Alien, telescope spot size matters only if you have seeing good enough to allow the telescope blur to dominate. Most of us have less than perfect seeing, though, and even if it is in the 1-1.5" range, seeing is going to be a more dominant factor than diffraction. With a big aperture, diffraction matters even less. The only time when diffraction and other telescope blur factors matter is with smaller telescopes with small apertures, where diffraction could be as large or larger than seeing.

 

With the RASA, since diffraction is not really of much concern, as long as you have good to great seeing, then yes, the small pixels should be of great help.



#7 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 23 November 2017 - 02:54 PM

Quick normalized noise comparison of 300s dark frames, between the Sony IMX183, Panasonic MN34230ALJ, and Sony ICX814. Reference basis was the STF for the ICX814. These cameras all had gains in the range of ~0.36e-/ADU to 0.48e-/ADU. Not identical, but close enough to minimize differences due to gain. Each camera is "oversampling" the electrons by over a 2:1 ratio, so the differences in bit depths don't really matter here. Quantization error is minuscule with all three cameras.

 

For these comparisons, I decided to "de-gap" the histograms of the CMOS cameras, as the gaps are not representative of the native camera/sensor behavior (conversion to 16-bit is purely a driver thing). To de-gap, I used the following pixel math:

 

K: $T/16

 

I further normalized the three images by balancing their means using the following pixel math:

 

K: $T + (mean(ICX814) - mean($T))

 

This converted the following histogram:

 

ERK8XNG.jpg

 

Into this histogram:

 

mFxrVAL.jpg

 

Removal of the gaps presents the true output of the sensor at the given gain. In comparison to the IMX183 histogram above, the MN34230ALJ looks like the following:

 

xFFy1T8.jpg

 

For reference, the ICX814 looks like the following:

 

5oEdnjN.jpg

 

The primary difference between the cameras here is read noise, which was about 1.55e- for the IMX183, 1.3e- for the MN34230ALJ, and around 3.5-4e- for the ICX814 (consider the square of each, 2.4e-, 1.7e- and 12.3e-...and the above comparison will probably make more sense). Secondarily, there may be an added noise discrepancy between the MN34230ALJ/ICX814 and the IMX183 due to dark current & glows...my measured dark signal with PI's BasicCCDParameters script (which I am not sure is accurately measuring dark current for these cameras) for the IMX183 was about an order of magnitude higher than the expected 0.002e-/s. This comparison should give you a rough idea of the previously mentioned noise difference between the ASI1600 and the ASI183. Note that this is a per-pixel noise comparison.


Edited by Jon Rista, 23 November 2017 - 04:04 PM.

  • h2ologg, n2068dd and mosentos like this

#8 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 23 November 2017 - 03:41 PM

My initial high-gain narrow band experiments with the ASI183MM Pro proved to be less than ideal. I started with Gain 200, as at the time I hadn't done any testing and had no charts from ZWO about what the gain settings were and how much noise there was at each. In the end, it turned out that Gain 200 was a bit higher gain in literal terms than on the ASI1600. It was 0.36e-/ADU, compared to 0.48e-/ADU for the ASI1600. Not a huge difference in terms of electron sampling, and not even all that large of a difference in read noise (~1.55e- vs. 1.3e-), however the ASI183 had less DR than the ASI1600. With the higher read noise and the brighter amp glows, that proved to be a bit too high of a gain.

 

For all of the images I'll be sharing here, I am performing the most bare-bones processing possible. This has sort of been my MO lately anyway, however I'm going even more bare bones here in an effort to demonstrate the true capabilities of this camera, without any noise reduction or deconvolution or star reduction getting in the way of an accurate and faithful presentation of what the hardware can do. I'll be sharing crops at 100%, as well as "binned 2x2", which in this case is just 50% downsampling in PI in it's auto mode. Additionally, I am not performing any linear fit on the data. The only alignment and color calibration involved here will be a simple "linear alignment", performed with the following pixel math:

 

K: $T + (mean(<ref>) - mean($T)

 

I may swap mean for median, depending on which gives better alignment of the background. In some cases, I may align the mean to the median or the median to the mean, again for the same purpose. In any case, a linear alignment only shifts the data positive or negative, it does not rescale nor redistribute tones. I have found it is a very effective way of producing very natural results with narrow band images. The only other processing I may do is DBE to correct gradients, and of course basic stretching. Hopefully this will present a clear picture of the noise this camera has, and how well (or not) signal buries that noise at different gain settings and with different integration times. If I was able to acquire any data other than Ha, I'll share a color image along with at least the grayscale Ha. I may also have LRGB, HaRGB, LHaRGB and even LHaRGBOIII images in the future, particularly as we move into galaxy season.

 

Also note that all images I will be presenting here unless otherwise specified (and this only goes for me, other people are welcome to share if they get one of these cameras in the future) are acquired with a Canon 600mm f/4 L II telephoto lens, on a belt modded Orion Atlas. This produces an image scale of ~0.82"/px with the IMX183. It's a good image scale, samples at around 2x with good seeing, and 3x or more with poorer seeing. It is a bit shy of ideal sampling for 2x seeing, though, which IMO would be 0.6"/px, and if you have even better seeing, sampling around 0.5"/px may be even better.

 

Elephant Trunk (#1)

 

Nevertheless, I acquired a few images at higher gain settings. The first of them was Elephant Trunk, which unintentionally also happened to be my first image with the ASI1600 at almost the same time last year. lol.gif Guess it was just an obvious target for new camera testing. Anyway, seeing was not terribly great, and my measured FWHMs were about the same with the ASI183 as with the ASI1600, around 2.4". The shorter subs did not help in the face of the average (or maybe even a bit worse than average, at least for here) seeing, and SNR was not as good as I had hoped. This was far less of a problem for Ha, of course, but was more of a problem for OIII, which delivered very little signal in even more time than I spent on Ha (which was about 2h15m for each channel).

 

The Ha channel was exposed fairly well on this moderately bright object in just 2.2 hours of integration, 44x180s subs (after culling about 10%):

 

6Vx7s3P.jpg

 

Contrast is pretty good, however SNR in the lower two corners was fairly poor. Longer subs at a lower gain would have (and have since proven to be) more sufficient for maintaining good SNR across the field.

 

The OIII channel was exposed slightly longer than the Ha, at 2.4 hours of integration, 48x180s subs (again, after culling about 10%). This produced a weak OIII signal, and this camera actually has fairly good OIII sensitivity (a fair bit better than the ASI1600, actually, with a peak over 70%). Regardless, to get a good contrasty OIII signal on this object, I believe an integration about 4x as deep would be required (so about 9 hours). The OIII signal was sufficient to highlight, slightly, some of the edges of the Ha structure, both in the lower part of the trunk and the lower right, bottom center region of the image:

 

wtp4FJU.jpg

 

For reference, on a calibrated wide gamut screen, the color of the image above is a slightly pinkish red. I have done no color processing here to correct for sRGB color space or anything like that, although the image has been saved in and is tagged as sRGB. If your screen is not well calibrated, the color may shift. In particular, poorly calibrated sRGB screens often render the color more of an orange-red, rather than the appropriate pinkish-magenta red. Contrast may also suffer.


Edited by Jon Rista, 23 November 2017 - 03:48 PM.

  • dswtan, h2ologg, n2068dd and 2 others like this

#9 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 23 November 2017 - 03:43 PM

With lower integration times, binning is always an option to improve SNR. While the >2x improvement in SNR that is possible with CCD cameras is not an option with CMOS, binning, or downsampling by 2x as it usually ends up being, will still improve pixel SNR by 2x. This is demonstrated here, producing a 5mp image out of the original 20mp size, and the results for about 2h15m of integration are still quite decent:

 

8cZXyyk.jpg


Edited by Jon Rista, 23 November 2017 - 03:44 PM.

  • dswtan, n2068dd, calypsob and 1 other like this

#10 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 23 November 2017 - 03:45 PM

Finally, a full size 100% crop showing the true noise characteristic of the image, which totals about 4h30m of data across both channels:

 

dZqgg3U.jpg

 

For a bit over 2 hours a channel, which is definitely within reach of beginners, it's a vastly superior result to what you might get with an OSC camera in any light polluted zone, and even more superior to any uncooled DSLR in an LP zone. Also keep in mind that this data has had zero noise reduction, so the results could be improved with careful NR.


Edited by Jon Rista, 23 November 2017 - 03:46 PM.

  • dswtan, h2ologg, n2068dd and 2 others like this

#11 evan9162

evan9162

    Apollo

  • *****
  • Posts: 1384
  • Joined: 20 Sep 2013
  • Loc: Boise, ID

Posted 24 November 2017 - 12:33 AM

Are these values with your 600 F/4L at F/4?



#12 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 November 2017 - 01:29 AM

Are these values with your 600 F/4L at F/4?

Yes, all at f/4.



#13 Henry from NZ

Henry from NZ

    Viking 1

  • *****
  • Posts: 958
  • Joined: 26 Jan 2014

Posted 24 November 2017 - 02:44 AM

Thanks Jon for sharing!

I would be most interested to see sample galaxy images taken with 130mm class refractor with this camera, which would be the main reason I consider this camera.

Also, what is the difference in efficiency between these two combo:

Imx183 2.4micron pixel at 800 mm f/l

Vs

A hypothetical camera 4.8 micron pixel at 1600 mm f/l

Assuming same f/ratio of f/6 for argument sake.

Which one needs shorter subs / less integration time and by how much?

Thanks

Edited by Henry from NZ, 24 November 2017 - 02:45 AM.

  • AstroCatinfo likes this

#14 AtmosFearIC

AtmosFearIC

    Viking 1

  • -----
  • Posts: 711
  • Joined: 10 Dec 2015
  • Loc: Melbourne

Posted 24 November 2017 - 03:39 AM

What you're looking at there is a 5" F/6 vs a 10" F/6 with a scaling amount of micron size.

With both of these systems you'll be imaging with the same imaging scale of ~0.62"/pixel but the 10" F/6 will have 4x the light gathering power. Exposure times will be the same but you'll get a stronger signal with the 10" F/6.


  • Rickster and Pug like this

#15 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 November 2017 - 03:44 AM

What you're looking at there is a 5" F/6 vs a 10" F/6 with a scaling amount of micron size.

With both of these systems you'll be imaging with the same imaging scale of ~0.62"/pixel but the 10" F/6 will have 4x the light gathering power. Exposure times will be the same but you'll get a stronger signal with the 10" F/6.

Only the star signals will be stronger with the 10". Extended object signals will be the same with both systems. 


  • Chris Appleton likes this

#16 Henry from NZ

Henry from NZ

    Viking 1

  • *****
  • Posts: 958
  • Joined: 26 Jan 2014

Posted 24 November 2017 - 07:07 AM

Jon, can you please elaborate? Are you saying that for galaxies or planetary nebulae the sub length and total integration time would be same for both set ups for a certain level of SNR?
  • Pug likes this

#17 Pauls72

Pauls72

    Surveyor 1

  • *****
  • Posts: 1961
  • Joined: 28 Oct 2007
  • Loc: LaPorte, IN USA

Posted 24 November 2017 - 09:47 AM

These are with a QHY183C OSC on a William Optics 98mm with a f0.8x reducer/flattener using 600 sec subs. Knowing what I do now, if I re-image them it will be with more subs of a shorter length.

Scope is f/6.3 @ fl 618mm with the reducer it is f/5 @ fl 494mm.

 

get.jpg?insecure

 

get.jpg?insecure


  • h2ologg, Jon Rista and Pug like this

#18 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 November 2017 - 12:29 PM

Jon, can you please elaborate? Are you saying that for galaxies or planetary nebulae the sub length and total integration time would be same for both set ups for a certain level of SNR?

Well first things first...in a non-normalized context, for extended objects, f-ratio determines exposure time. If two scopes are f/6, then no matter what their other characteristics...doesn't matter what each one's aperture and focal length are...then both will require the same exposure length to put the same number of photons into a given unit area on the sensor. 

 

You either have a shorter focal length with a smaller aperture, or a longer focal length with a larger aperture. For extended objects, the *image* resolved by the scope is projected onto the sensor. As that image moves down the focal length, it is magnified...the light is spread out. So while a larger aperture gathers more light, if that light has to travel a greater distance, the larger quantity of light falls off more than if it travelled a shorter distance. Conversely, the smaller aperture of the shorter scope gathers less light, but that light travels over a shorter distance, so it falls off less. This is why f-ratio is such a useful thing. It allows us to gauge exposure, quickly an easily, regardless of how long or big a scope is.

 

That is for extended objects. For stars, point sources (which are defined as 1/4 the angular size of the airy disc), where all of the energy of the object is contained within the airy pattern, things are a little different. Both f-ratio and aperture affect the exposure of point sources. So at a given f-ratio, a larger aperture will mean you can resolve smaller stars, and all stars will expose to a brighter level in a given amount of exposure time.

 

In your example setups, you actually normalized the image scale:

 

A) 800mm fl, 134mm aperture, f/6, 2.4 micron, 0.62"/px

B) 1600mm fl, 267mm aperture, f/6, 4.8 micron, 0.62"/px

 

These two systems have identical exposure and (pixel) SNR characteristics, assuming read noise scales as well. So a 10 minute sub with the 800mm scope will produce the same exposure as a 10 minute sub with the 1600mm scope.

 

There will be only two key differences between these scopes. The first is FoV, the longer scope sees a smaller angular area of the sky. The object is magnified more, it's light is spread out more. If you used the same 2.4 micron pixels at 1600mm, then the pixel SNR of each 2.4 micron pixel would be less. You would have to bin 2x2 in order to normalize the pixel SNR of the 1600mm setup with that of the 800mm setup if you used 2.4 micron pixels with both. But since you already specified 4.8 micron pixels, then the 1600mm setup will deliver the same pixel SNR in the same exposure time. It will just render the object larger within the field. 

 

The other key difference is that the 1600mm setup will have stars that are twice as bright. Because the aperture is gathering more light in total, it is gathering more light on stars as well. The difference with stars is that since they are effectively a mathematical point, all of their light is focused into the airy pattern on the sensor. Until such a time as your airy pattern is magnified so much that it is significantly larger than a pixel, the bigger the aperture, the brighter the stars. So if you have two scopes with the same f-ratio, but different apertures, the one with the larger aperture will resolve more and smaller stars. 


Edited by Jon Rista, 24 November 2017 - 07:40 PM.

  • h2ologg, Pauls72, scopenitout and 2 others like this

#19 Henry from NZ

Henry from NZ

    Viking 1

  • *****
  • Posts: 958
  • Joined: 26 Jan 2014

Posted 24 November 2017 - 01:45 PM

Thank you Jon for the detailed explanation.
Since when I am doing AP I am more interested in the nebulously or galaxy structures rather than the stars themselves, I suppose that means that there is little point in cranking up the focal length/aperture (for same f/ ratio).

That begs the question, is there still any merit having large scopes like 10” or 12” RC if one really just want to image small fuzzies? Large aperture scope setups are much more expensive than a new small pixel camera. This is assuming that the small pixel cameras has same sensitivity as larger pixel cameras of course.

For example a 10” Truss RC reduced to f/6-7 will set me back about $4K in local currency, whereas a new 2.4 micron camera will be about $2k and I can use my existing 5” refractor to image at the same f/ratio with similar arcsec/px.

Will there still be a market in the future for big aperture scopes for DSO AP??

Edited by Henry from NZ, 24 November 2017 - 01:55 PM.

  • AstroCatinfo and Pug like this

#20 *Axel*

*Axel*

    Vostok 1

  • -----
  • Posts: 107
  • Joined: 01 Mar 2016

Posted 24 November 2017 - 01:58 PM

Thank you Jon for the detailed explanation.
Since when I am doing AP I am more interested in the nebulously or galaxy structures rather than the stars themselves, I suppose that means that there is little point in cranking up the focal length/aperture (for same f/ ratio).

That begs the question, is there still any merit having large scopes like 10” or 12” RC if one really just want to image small fuzzies? Large aperture scope setups are much more expensive than a new small pixel camera. This is assuming that the small pixel cameras has same sensitivity as larger pixel cameras of course.

For example a 10” Truss RC reduced to f/6-7 will set me back about $4K in local currency, whereas a new 2.4 micron camera will be about $2k and I can use my existing 5” refractor to image at the same f/ratio with similar arcsec/px

Henry,

 

Even though it doesn't matter resolution wise  (for ex. here 0.62"/px as stated above) FL  still matters with respect to FOV knowing that there is a high count of NGCs below 30' in size especially galaxies .


  • Jon Rista and Pug like this

#21 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 November 2017 - 02:47 PM

Thank you Jon for the detailed explanation.
Since when I am doing AP I am more interested in the nebulously or galaxy structures rather than the stars themselves, I suppose that means that there is little point in cranking up the focal length/aperture (for same f/ ratio).

That begs the question, is there still any merit having large scopes like 10” or 12” RC if one really just want to image small fuzzies? Large aperture scope setups are much more expensive than a new small pixel camera. This is assuming that the small pixel cameras has same sensitivity as larger pixel cameras of course.

For example a 10” Truss RC reduced to f/6-7 will set me back about $4K in local currency, whereas a new 2.4 micron camera will be about $2k and I can use my existing 5” refractor to image at the same f/ratio with similar arcsec/px.

Will there still be a market in the future for big aperture scopes for DSO AP??

Well, at the same f-ratio, a larger aperture means longer focal length. That means higher resolution, up to a point. If your seeing is good, then a larger scope will be able to resolve finer details (that goes for both stars as well as extended object features). That is the main benefit of a larger aperture on extended objects: resolution. However, if you DO NOT have the seeing for it, then a larger aperture may not buy you anything, other than possibly a more frustrating experience as you need better tracking and better guiding for higher resolution results.

 

Additionally, if you use larger pixels with the longer scope, then you aren't really going to be resolving more detail in the end...the pixels will limit your resolution. This is actually the most intriguing thing about the IMX183...because it has such small pixels, it presents the opportunity to get good resolution with smaller scopes that are less demanding on the mount. While image scale is the same with smaller pixels on the shorter focal length, the smaller scope will usually weigh less, so it should be easier to get good tracking with it. You can also get scopes that require less fiddling...say a refractor. A newt is also usually going to be easier to collimate than an RC. An SCT, even reduced, is going to result in a much larger image scale regardless, so they will usually put greater demands on tracking unless it is a really small SCT.

 

Regarding aperture and stars. This actually has interesting implications for dynamic range. If you keep f-ratio the same but increase aperture, then your stars will be brighter in the same exposure time. That has a NEGATIVE consequence with dynamic range...same exposure, brighter stars...you need more dynamic range. Or alternatively, you end up clipping your stars more quickly with a larger aperture at the same f-ratio. (This is actually a problem I have with my 600mm f/4, which is a 150mm aperture...while many people image at around 500-700mm focal lengths with shorter refractors, not many are at f/4...usually it's around f/5-f/6.8 or so, apertures around 80-100mm...a smaller aperture. That means they will have fewer problems with dynamic range and clipped stars than I do, in the same exposure time, and even with longer exposures, they will usually not have the same problems with clipped stars. Their only consequence may be lower resolution, and if seeing is good, the larger diffraction might reduce resolution by a small amount.)

 

FTR, with the IMX183 on my 600mm lens, I have fewer problems with clipped stars than I did with the Panasonic MN34230ALJ on the same lens. Stars are sampled better, so more starlight with average seeing (which at best guess is around 2" where I live, with some nights being closer to 3" and some nights being closer to 1.3-1.5") is spread around a few pixels. With less star intensity per pixel, and with the higher dynamic range on the IMX183 sensor @ Gain 53, I have actually had fewer problems with clipped stars than on the Panasonic. Since I need longer exposures, that has been a real bonus, as it's actually allowed me to get the necessary longer subs to properly swamp read noise. Conversely, with the Panasonic, because the read noise is so low at high gain, I don't need exposures as long. I figure if I had 11.5 stops of DR rather than 10.4 stops of DR at Gain 200, I wouldn't have any clipped stars on the Panasonic. Well, you make do with what you have. wink.gif

 

Oh, one final note. The ASI183MM Pro has definitely been a bit more challenging to use than the ASI1600MM Cool. While in general I've been working on improving my focusing over the last year, once I started using the ASI183MM Pro I really had to step up my focus game. The smaller pixels reveal defocus much more readily, and I have to really keep on top of focus to maintain good resolution. I have even resorted to partial manual control, to keep the lens focused without having to spend several minutes every couple of frames running the SGP AF routine. In part this is because my lens doesn't seem to cool in an entirely uniform way, so using temperature compensation has not been as effective as I had hoped, and depending on the time of night and relative equalization of the scope, the compensation may even result in worse problems. A telescope with simpler optical design may not quite have the same issues. In any case, keeping on top of focus at all times has been necessary to achieve good resolution with the ASI183. And I've not always been successful. Additionally, the moment seeing goes from good or great to average or poor, the impact to resolution is much more readily observable in my subs than it ever was with the ASI1600 (which itself is actually a fairly high resolution sensor with relatively small pixels!) Where I could handle a guide RMS of around 0.75-0.8" with the ASI1600, I really need no worse than 0.6" with the ASI183. I've seen continued benefits with guide RMS down to 0.45" with the ASI183 when I am at ideal focus as well. There have been nights where I've scrapped, or at least marked as bad, an entire nights worth of subs because seeing was so bad, I couldn't really use the data...not, at least, if I wanted high resolution details.


Edited by Jon Rista, 24 November 2017 - 03:09 PM.

  • h2ologg, AstroCatinfo and Pug like this

#22 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 November 2017 - 03:56 PM

My second high gain image with the ASI183MM Pro ended up being a chain of mishaps, mixed with a few nights of absolutely horrific seeing. I made two mistakes with this image. The first was using Gain 200, which as with my first image of Elephant Trunk, resulted in too little dynamic range. I was unable to swamp the read noise enough, only around 2-3xRN^2, which for most subs was below my lower cutoff limit of 3xRN^2 (I recommend this be the bare minimum you aim to swamp read noise by with any camera.) Interestingly, this was despite an increase in exposure length from the 180s I used on Trunk to 210s. I didn't measure the skies these nights, however based on my background sky levels, it seems the skies were much darker than I've become accustomed to, so the longer exposures resulted in even less swamping of the noise as with the other targets that were using 180s subs.

 

My second mistake, and I don't know if it was user error, or a glitch in SGP, I somehow ended up with about half the subs UNDITHERED. I noticed near the end of one of the nights I images that dithering was disabled. After checking subs from a few nights, it had apparently been disabled for several nights. This seems to have had a fairly significant impact on my final SNR, since the FPN wasn't able to be average out for about half the stack. This issue was compounded by the fact that I ended up having to toss about two nights worth of extra data I acquired to try and replace these undithered frames due to poor seeing and some focus slippage issues. (Yeah, this image was kind of a disaster!!) I culled over 30% of the subs from this stack, AFTER having previously discarded a number of bad subs within SGP due to tracking issues. (Double disaster. I always seem to run into problems when trying to image Cave!! I love the object, but have as of yet been unable to acquire a full set of NB data across all three channels on it...well, I guess next year!)

 

I acquired over 11 hours of data, and stacked 8h17m here (147x210s). A lot of the discards were subs that had over 5" FWHMs, which absolutely decimated any details and cost me about half the stars that were visible with subs that had FWHMs under 3":

 

aNeygDW.gif

 

Up to about 2.5" or so (and even up to 3") the loss in resolution wasn't bad enough for me to discard the subs, and I figured I could just weight them with SubframeSelector in PI to optimize SNR and resolution. But beyond 2.5-3" and things were just too blurry (remember, the image scale is 0.82"/px here). The fact that I had nights with 4" or worse seeing was somewhat ironic as I also measured some of my best FWHMs on some of the nights I imaged this object, down to around 1.7", which occurred on nights with some of the best seeing I've had all year:

 

3BNPUIx.jpg

 

Such is the nature of our hobby. Always at the mercy of the environment. Boy, what I would give to have 1.5" seeing all the time!! ;)


  • Pug likes this

#23 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 November 2017 - 04:04 PM

Cave Nebula (#2)

 

This is one of my favorite dark and dusty regions up in the night sky. I really love images of this region, especially LRGB images from dark sites. It's such a colorful region with so much detail. I've been trying to image Cave since I first started imaging. I had a couple attempts with DSLR &  LP filters in my back yard, and the data was totally unusable...barely revealed the brightest portions of cave at all. I tried once at my dark site with the 5D III, but we had some kind of atmospheric phenomena that night, and both the Cave nebula as well as Flaming Star nebula ended up quite washed out behind a very bright silvery haze. Never figured out what happened those nights. I tried to image Cave again last year with the ASI1600...it was one of the first images I made with that camera last year. I again had issues, and was again only able to get a small amount of integration just in Ha.

 

Here is an example of the full field at 600mm focal length:

 

1UqQIgB.jpg

 

Dark and dusty regions like this demand deeper exposures, as so much more of the image ends up close to the read noise threshold. Technically speaking, at a true dark site, narrow band images of a region like this would never be able to be exposed according to any exposure rule. The darkest regions would remain read noise limited, as there simply isn't enough skyglow at a dark site to produce enough signal to swamp the noise unless you expose for hours (which is just not an option with higher gain, lower noise cameras.) So gauging proper exposure with narrow band for objects like this is a bit of a challenge regardless. FTR, I used 3nm AstroDon filters here, and I estimate my skies were between 18.7 and 18.9 mag/sq".

 

Here is a 2x2 binned crop of the core nebula itself:

 

zIUSzZr.jpg

 

SNR is not all that great here, however considering all the issues I had when imaging, it's not so bad. The noise with this camera takes on a very pleasing characteristic when calibrated. Ideally, I should have used about 7.5 minute subs at Gain 53, rather than 3.5 minute subs at Gain 200, to get better SNR. I ultimately ended up using 10 minute subs at Gain 53 to swamp the read noise by more than 3xRN^2, and to compensate a bit for the extra noise from amp glow. (I'll share some integrations with 10 minute subs soon.)


Edited by Jon Rista, 24 November 2017 - 04:24 PM.

  • calypsob, David-LR and dvalid like this

#24 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 November 2017 - 04:10 PM

Here are a couple 100% crops of the interesting detail areas:

 

O454SJw.jpg

 

xYcx3Su.jpg

 

My recommendation when using this camera for higher resolution imaging is, avoid high gain unless you are purposely doing "lucky" imaging or very short exposure imaging (10s or shorter subs.) Use a moderate gain, either unity (Gain 120) or Gain 53 (which is about half unity, ~2e-/ADU gain). Additionally, make sure you dither. Dithering is very important in general, IMO. Just because the noise may appear to be spatially random does not mean it is also temporally random. Spatially random noise that is fixed in time is FPN, and it can and will limit your SNR if you do not dither it.


Edited by Jon Rista, 24 November 2017 - 04:12 PM.


#25 Jon Rista

Jon Rista

    ISS

  • *****
  • topic starter
  • Posts: 20228
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 November 2017 - 04:16 PM

For those interested, here is a comparison of the FoV difference between the ASI1600 and ASI183:

 

ASI1600:

6zWLO6L.jpg

 

ASI183:

1UqQIgB.jpg


  • h2ologg, m1618, David-LR and 7 others like this


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics